Method and system for predicting onset risk of type 2 diabetic nephropathy

Abstract

The invention relates to the technical field of diabetic nephropathy onset risks, in particular to a method and system for predicting the onset risk of type 2 diabetic nephropathy, and the method comprises the following steps: screening out characteristic factors having significant influence on the onset of type 2 diabetic nephropathy by using a multi-factor Logistic regression model; based on the contribution degree of the feature factor, determining a feature weight used for performing centralization processing on the feature factor, and determining a feature weight used for performing decentralization processing on the feature factor; gaming is carried out between the feature weight after decentralization processing and the feature weight after centralization processing, and an optimal feature weight is obtained; and on the basis of the feature factors weighted by the optimal feature weight, constructing an onset risk prediction model of the type 2 diabetic nephropathy. According to the method, the prediction model capable of efficiently identifying high-risk crowds is constructed, the onset risk of the type 2 diabetic nephropathy of a patient is predicted through characteristic factors, and objective, accurate and standard-unified risk prediction is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of diabetic nephropathy risk prediction, and in particular to a method and system for predicting the risk of type 2 diabetic nephropathy. Background Art

[0002] Type 2 diabetic nephropathy (DKD) is one of the most common complications of diabetes patients and the main cause of end-stage renal disease (ESRD), accounting for 59% of new ESRD. With the increasing incidence of diabetes year by year, the high incidence of DKD has brought tremendous pressure and challenges to the medical system and social economy.

[0003] However, there is currently a lack of DKD epidemiological research based on national data and the establishment of automated prediction models. Most early DKD risk assessments still rely on the subjective judgment of doctors' personal experience, which is prone to subjective bias. Therefore, it is impossible to achieve objective and accurate prediction of early DKD risk. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for predicting the risk of developing type 2 diabetic nephropathy, so as to solve the technical problem that the early risk assessment of DKD in the prior art mostly relies on the subjective judgment of the doctor's personal experience, is prone to subjective bias, and cannot achieve objective and accurate prediction of the early risk of DKD.

[0005] In order to solve the above technical problems, the present invention specifically provides the following technical solutions: A method for predicting the risk of developing type 2 diabetic nephropathy comprises the following steps: The multivariate Logistic regression model was used to analyze the influencing factors of type 2 diabetic nephropathy and screen out the characteristic factors that have a significant impact on the onset of type 2 diabetic nephropathy. Conduct principal component analysis on each characteristic factor to obtain the contribution of the characteristic factor; Based on the contribution of the characteristic factors, determining characteristic weights for centralizing the characteristic factors; Determining feature weights for decentralizing the feature factors based on the contribution of the feature factors; A game is played between the feature weights after decentralized processing and the feature weights after centralized processing to obtain the optimal feature weights; Based on the characteristic factors weighted by the optimal characteristic weights, a risk prediction model for type 2 diabetic nephropathy was constructed.

[0006] As a preferred embodiment of the present invention, the influencing factors include: demographic characteristics, clinical information, laboratory tests, renal puncture pathology results, physical examination, lifestyle, drug use, past medical history and family history, socioeconomic status, and emotional state.

[0007] As a preferred embodiment of the present invention, the method for screening the characteristic factors comprises: SPSS26.0 and RStudio4.2.2 software were used to analyze the case data of type 2 diabetic nephropathy. The quantitative data that conformed to the normal distribution were expressed as mean ± standard deviation, and the qualitative data that did not conform to the normal distribution were expressed as frequency and percentage. The differences in influencing factors between patients without type 2 diabetic nephropathy and patients with type 2 diabetic nephropathy were compared. The t-test was used to compare the quantitative data with normal distribution between the groups, and the chi-square test and rank sum test were used to compare the qualitative data that did not conform to the normal distribution between the groups. The difference was statistically significant; Logistic regression analysis was used to screen the influencing factors, with the test level α=0.05, and the screened influencing factors were used as characteristic factors.

[0008] As a preferred solution of the present invention, a method for determining feature weights after centralization processing includes: The contribution of the characteristic factor is normalized and used as the characteristic weight after the characteristic factor is centralized. The normalization process of the feature weights after the centralization process is: ; In the formula, After centralized processing i The feature weight of each feature factor, For the i The contribution of the characteristic factors, is the contribution set of each characteristic factor, for The minimum value in The maximum value in .

[0009] As a preferred solution of the present invention, a method for determining feature weights after decentralized processing includes: The contribution of the characteristic factor is processed inversely, and the contribution after the inverse processing is normalized as the characteristic weight of the characteristic factor after decentralized processing; The inverse processing and normalization processing of the feature weight after the centralization processing are: ; In the formula, After decentralization i The feature weight of each feature factor, For the iThe contribution of the characteristic factors, is the inverse set of contribution of each characteristic factor, The maximum value in , n is the total number of feature factors, min is the minimum operator, and max is the maximum operator. As a preferred solution of the present invention, the method for determining the optimal feature weight includes: The game theory centralized model is used to optimize the feature weights after decentralized processing and the feature weights after centralized processing. The decision model of the optimal feature weights is derived from the optimization goal of minimizing the weight deviation. The decision model of the optimal feature weights is: ; In the formula, F is the objective function value, min is the minimum operator, x is the identifier of w and v, for The linear coefficient of , T is the transposition operator; The decision model is transformed and solved using the optimal first-order derivative condition, and the solution equation for the optimal feature weight is obtained as follows: ; Solve the solution equation and normalize it to obtain the optimal feature weight, which is: ; ; In the formula, is the set of optimal feature weights of each feature factor, After normalization , is a set of linear coefficients, for The minimum value in The maximum value in .

[0010] As a preferred embodiment of the present invention, the method for constructing the risk prediction model for type 2 diabetic nephropathy comprises: Construct training and test sets based on case data of type 2 diabetic nephropathy; The feature factors weighted by the optimal feature weights are used as the input of the convolutional neural network, and the risk score is used as the output of the convolutional neural network; The convolutional neural network was trained using the training set to obtain a risk prediction model for type 2 diabetic nephropathy, where the risk prediction model for type 2 diabetic nephropathy is: ; In the formula, Score is the risk score, are the optimal feature weights of the 1st, 2nd, and nth feature factors respectively, They are the 1st, 2nd, and nth feature factors respectively, and CNN is a convolutional neural network; The test set was used to evaluate the performance of the risk prediction model for type 2 diabetic nephropathy.

[0011] As a preferred solution of the present invention, in the process of training the convolutional neural network, a multi-dimensional cross-validation method is used to optimize the hyperparameters of the convolutional neural network; In the performance evaluation of the risk prediction model for type 2 diabetic nephropathy, the calibration curve and D'Agostino-Nam test were used to evaluate the calibration of the model, and the area under the ROC curve and Harrell's C index were used to evaluate the discrimination ability of the model.

[0012] As a preferred embodiment of the present invention, the risk level determination method includes: classifying the risk score range of the top 25% of all cases in the case data of type 2 diabetic nephropathy as the threshold range of the high risk level, the middle 50% of the risk score range as the threshold range of the medium risk level, and the bottom 25% of the risk score range as the threshold range of the low risk level.

[0013] As a preferred embodiment of the present invention, the present invention provides a risk prediction system for type 2 diabetic nephropathy, which is applied to a method for predicting the risk of type 2 diabetic nephropathy. The system comprises: The feature selection unit is used to analyze the influence degree of the factors affecting the onset of type 2 diabetic nephropathy using a multi-factor Logistic regression model, and to screen out the characteristic factors that have a significant impact on the onset of type 2 diabetic nephropathy; A feature preprocessing unit is used to perform principal component analysis on each feature factor to obtain the contribution of the feature factor; based on the contribution of the feature factor, determine the feature weight for centralizing the feature factor; based on the contribution of the feature factor, determine the feature weight for decentralizing the feature factor; and perform a game between the feature weight after decentralized processing and the feature weight after centralized processing to obtain the optimal feature weight; The deep learning unit is used to construct a risk prediction model for type 2 diabetic nephropathy based on feature factors weighted by optimal feature weights.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention utilizes a neural network to construct a prediction model that can efficiently identify high-risk populations, thereby predicting the risk of type 2 diabetic nephropathy in patients by obtaining characteristic factors from clinical data, thereby achieving objective, accurate and standardized risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0016] Figure 1 A flow chart of a method for predicting the risk of developing type 2 diabetic nephropathy provided in an embodiment of the present invention; Figure 2 A system framework for predicting the risk of type 2 diabetic nephropathy provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] like Figure 1 As shown, the present invention provides a method for predicting the risk of developing type 2 diabetic nephropathy, comprising the following steps: The multivariate Logistic regression model was used to analyze the influencing factors of type 2 diabetic nephropathy and screen out the characteristic factors that have a significant impact on the onset of type 2 diabetic nephropathy. Conduct principal component analysis on each characteristic factor to obtain the contribution of the characteristic factor; Based on the contribution of the characteristic factors, the characteristic weights used to centralize the characteristic factors are determined; Based on the contribution of the characteristic factors, determining the characteristic weights for decentralizing the characteristic factors; A game is played between the feature weights after decentralized processing and the feature weights after centralized processing to obtain the optimal feature weights; Based on the characteristic factors weighted by the optimal characteristic weights, a risk prediction model for type 2 diabetic nephropathy was constructed.

[0019] The present invention firstly collects all the influencing factors related to the onset of type 2 diabetic nephropathy, and uses statistical methods to analyze the degree of influence using a multivariate logistic regression model, so as to screen out factors that have a significant impact on the development of type 2 diabetic nephropathy in type 2 diabetic patients from various influencing factors, and provide scientific feature selection for the subsequent construction of a disease risk prediction model, that is, to construct a disease risk prediction model that calculates the disease risk based on characteristic factors.

[0020] After the present invention screens the various characteristic factors, principal component analysis is performed on each characteristic factor to identify the contribution of each characteristic factor to the risk of disease. The higher the contribution, the more important the characteristic factor is, and the disease risk prediction model will be more biased towards the information provided by the characteristic factor in the process of disease risk calculation.

[0021] The present invention adopts two ways of assigning weights to characteristic factors. The first is that the higher the contribution of the characteristic factor, the higher the characteristic weight assigned to the corresponding characteristic factor. The weighting of the characteristic weight is used to further enhance the importance of the characteristic factor, so that the disease risk prediction model will further tend to the information provided by the characteristic factors with high importance during the disease risk calculation process, thereby causing the disease risk prediction model to focus on the characteristics with high importance (containing more information related to the disease risk calculation) when processing each characteristic factor, and ignore the characteristics with low importance (containing less information related to the disease risk calculation), so as to achieve the effect of extracting relevant information and eliminating redundant information, and achieve the centralization of characteristic information. The disease risk prediction model of type 2 diabetic nephropathy trained in this way processes less characteristic information, improves efficiency and performance, and is not prone to overfitting defects.

[0022] The second is that the lower the contribution of the characteristic factor, the higher the characteristic weight is assigned to the corresponding characteristic factor. The importance of the characteristic factor is weakened by weighting the characteristic weight, so that the disease risk prediction model will be biased towards the information provided by the characteristic factors with low importance during the disease risk calculation process. As a result, when processing each characteristic factor, the disease risk prediction model will shift its attention from the features with high importance (containing more information related to the disease risk calculation) to the features with low importance (although containing less information related to the disease risk calculation, it still contains a certain amount of information used for disease risk calculation), so as to achieve the effect of comprehensive information grasp and avoid the centralization of characteristic information. The disease risk prediction model of type 2 diabetic nephropathy trained in this way can process more characteristic information, improve the accuracy performance, and is not prone to underfitting defects.

[0023] The above two ways of assigning weights to characteristic factors have their own advantages and complement each other. Therefore, the present invention uses a game theory method to make the best decision between the characteristic weights obtained by the above two ways of assigning weights to characteristic factors, and obtains the optimal characteristic weights, so that the optimal characteristic weights decided can seek a compromise between their respective advantages and the advantages of each other, thereby achieving a balance between the two advantages. The risk prediction model for type 2 diabetic nephropathy trained by weighting the characteristic factors with the optimal characteristic weights takes into account the improvement of efficiency performance and accuracy performance, so that the risk prediction model for type 2 diabetic nephropathy has high efficiency performance while obtaining high accuracy performance.

[0024] The optimal feature weights of the present invention are used to train the risk prediction model for type 2 diabetic nephropathy by weighting the feature factors. This model can evaluate the corresponding risk of type 2 diabetic nephropathy according to the patient's feature factors and determine the risk level. This can provide dynamic monitoring, stratified management, personalized intervention and feedback support for the patients in the future, laying a foundation for helping them improve their self-management capabilities.

[0025] Influencing factors include: demographic characteristics, clinical information, laboratory tests, renal puncture pathology results, physical examination, lifestyle, medication use, past medical history and family history, socioeconomic status, and emotional state. Demographic characteristics included: name, ID number, gender, date of birth, and place of residence; Clinical information included: time of diabetes diagnosis, diabetic nephropathy screening, and the presence of diabetic retinopathy, diabetic peripheral neuropathy, and diabetic macroangiopathy; Laboratory tests include: fasting blood glucose, glycosylated hemoglobin, blood pressure, blood lipids, renal function indicators (such as eGFR, urine protein level), blood routine, urine routine; Renal biopsy pathological results include: immunofluorescence examination, light microscopy, and electron microscopy results; Physical examination: height and weight at admission; Lifestyle: eating habits, sleep, smoking, drinking, physical activity; Drug use: use of hypoglycemic drugs, antihypertensive drugs, lipid-lowering drugs, etc.; Past medical history and family history: history of hypertension, lipid metabolism disorders, coronary heart disease and cerebrovascular disease, family history of diabetes; Socioeconomic status: annual household income per capita, education level, work status, marital status, and type of medical reimbursement; Emotional state: depression, anxiety symptoms.

[0026] The screening methods of characteristic factors include: SPSS26.0 and RStudio4.2.2 software were used to analyze the case data of type 2 diabetic nephropathy. The quantitative data that met the normal distribution were expressed as mean ± standard deviation. It indicates that the qualitative data that do not conform to normal distribution are expressed as frequency and percentage (%); The differences in influencing factors between patients without type 2 diabetic nephropathy and patients with type 2 diabetic nephropathy were compared. The t-test was used to compare the quantitative data with normal distribution between the groups, and the chi-square test and rank sum test were used to compare the qualitative data that did not conform to the normal distribution between the groups. The difference was statistically significant; Logistic regression analysis was used to screen the influencing factors, with the test level α=0.05, and the screened influencing factors were used as characteristic factors.

[0027] The first method of assigning weights to characteristic factors adopted by the present invention is that the higher the contribution of the characteristic factor, the higher the characteristic weight assigned to the corresponding characteristic factor, and the weighting of the characteristic weight is used to further enhance the importance of the characteristic factor, so that the disease risk prediction model will further favor the information provided by the characteristic factors with high importance during the disease risk calculation process, thereby causing the disease risk prediction model to focus on the characteristics with high importance (including more information related to the disease risk calculation) when processing each characteristic factor, and ignore the characteristics with low importance (including less information related to the disease risk calculation), so as to achieve the effect of extracting relevant information and eliminating redundant information, and achieve the centralization of characteristic information. The disease risk prediction model of type 2 diabetic nephropathy trained in this way processes less characteristic information, improves efficiency and performance, and is not prone to overfitting defects, as follows: Methods for determining feature weights after centralization include: The contribution of the characteristic factor is normalized and used as the characteristic weight after the characteristic factor is centralized. The normalization of feature weights after centralization is: ; In the formula, After centralized processing i The feature weight of each feature factor, For the i The contribution of the characteristic factors, is the contribution set of each characteristic factor, for The minimum value in The maximum value in .

[0028] The second method of assigning weights to characteristic factors adopted by the present invention is that the lower the contribution of the characteristic factor, the higher the characteristic weight assigned to the corresponding characteristic factor, and the weighting of the characteristic weight is used to weaken the importance of the characteristic factor, so that the disease risk prediction model will be biased towards the information provided by the characteristic factors with low importance during the disease risk calculation process, thereby causing the disease risk prediction model to shift its attention from the characteristics with high importance (containing more information related to the disease risk calculation) to the characteristics with low importance (although containing less information related to the disease risk calculation, it still contains a certain amount of information used for the disease risk calculation) when processing each characteristic factor, so as to achieve the effect of comprehensive information mastery and avoid the centralization of characteristic information. The disease risk prediction model of type 2 diabetic nephropathy trained in this way processes more characteristic information, improves the accuracy performance, and is not prone to underfitting defects, as follows: Methods for determining feature weights after decentralized processing include: The contribution of the characteristic factor is processed inversely, and the contribution after the inverse processing is normalized as the characteristic weight of the characteristic factor after decentralized processing; The inverse and normalized processing of the feature weight after centralization is: ; In the formula, After decentralization i The feature weight of each feature factor, For the i The contribution of the characteristic factors, is the inverse set of contribution of each characteristic factor, The minimum value in The maximum value in , n is the total number of characteristic factors, min is the minimum operator, and max is the maximum operator. In the present invention, the reciprocal processing is used to make the characteristic factors with large contribution have small characteristic weights, and the characteristic factors with small contribution have large characteristic weights.

[0029] The above two ways of assigning weights to characteristic factors have their own advantages and complement each other. Therefore, the present invention uses a game method to make the best decision between the characteristic weights obtained by the above two ways of assigning weights to characteristic factors, and obtains the optimal characteristic weights, so that the optimal characteristic weights determined can seek a compromise between their respective advantages and the advantages of the other party, thereby taking both advantages into account. The risk prediction model for type 2 diabetic nephropathy trained by weighting the characteristic factors with the optimal characteristic weights takes into account the improvement of efficiency performance and accuracy performance, so that the risk prediction model for type 2 diabetic nephropathy has high efficiency performance while obtaining high accuracy performance, as follows: Methods for determining the optimal feature weights include: The game theory centralized model is used to optimize the feature weights after decentralized processing and the feature weights after centralized processing. The decision model of the optimal feature weights is derived from the optimization goal of minimizing the weight deviation. The decision model of the optimal feature weights is: ; In the formula, F is the objective function value, min is the minimum operator, x is the identifier of w and v, for The linear coefficient of , T is the transposition operator; The decision model is transformed and solved using the optimal first-order derivative condition, and the solution equation for the optimal feature weight is obtained as follows: ; Solve the equation and normalize it to get the optimal feature weight, which is: ; ; In the formula, is the set of optimal feature weights of each feature factor, After normalization , is a set of linear coefficients, for The minimum value in The maximum value in .

[0030] Among them, the core idea of ​​the game theory aggregation model is to seek consistency or compromise between the indicator weights obtained by different weighting methods, minimize the deviation between the possible weights and the basic weights, and improve the scientific nature of indicator weighting.

[0031] The optimal feature weights of the present invention are used to train the risk prediction model of type 2 diabetic nephropathy by weighting the feature factors. According to the patient's feature factors, the corresponding risk of type 2 diabetic nephropathy can be evaluated and the risk level can be determined, as follows: The method for constructing a risk prediction model for type 2 diabetic nephropathy includes: Construct training and test sets based on case data of type 2 diabetic nephropathy; The feature factors weighted by the optimal feature weights are used as the input of the convolutional neural network, and the risk score is used as the output of the convolutional neural network; The convolutional neural network was trained using the training set to obtain a risk prediction model for type 2 diabetic nephropathy. The risk prediction model for type 2 diabetic nephropathy is: ; In the formula, Score is the risk score, are the optimal feature weights of the 1st, 2nd, and nth feature factors respectively, They are the 1st, 2nd, and nth feature factors respectively, and CNN is a convolutional neural network; The test set was used to evaluate the performance of the risk prediction model for type 2 diabetic nephropathy.

[0032] In the process of training the convolutional neural network, a multi-dimensional cross-validation method is used to optimize the hyperparameters of the convolutional neural network; In the performance evaluation of the risk prediction model for type 2 diabetic nephropathy, the calibration curve and D'Agostino-Nam test were used to evaluate the calibration of the model, and the area under the ROC curve and Harrell's C index were used to evaluate the discrimination ability of the model.

[0033] The risk level determination methods include: The risk score range of the top 25% of all cases in the case data of type 2 diabetic nephropathy was classified as the threshold range of high risk level, the risk score range of the middle 50% was classified as the threshold range of medium risk level, and the risk score range of the bottom 25% was classified as the threshold range of low risk level.

[0034] The present invention collects case data and constructs training and test sets: Research design and subjects: Based on a prospective cohort study design, patients diagnosed with type 2 diabetes in Hospital A, Hospital B, and Hospital C between 2016 and 2018 were selected as research subjects. Inclusion criteria: ① Living in the province where the hospital is located and aged ≥ 18 years; ② Not diagnosed with diabetic nephropathy Exclusion criteria: ① Other serious diseases that may cause renal insufficiency; ② History of acute kidney injury during the study period or within 6 months before the start of the study. ③ Key baseline or follow-up data are missing in the medical records. Baseline: From 2016 to 2018, based on the risk factors identified in the first research content, the hospital information management platform, the health records of residents in the province where the hospital is located, and the chronic disease monitoring system data were integrated to collect and organize the relevant indicator information of the research subjects at the baseline.

[0035] Follow-up: After the completion of the baseline survey, follow-up will be conducted every 6 months until the study period is reached. The survey content is the same as the baseline content.

[0036] Study outcomes: Percutaneous renal biopsy confirmed type 2 diabetic nephropathy.

[0037] Study endpoints: development of type 2 diabetic nephropathy, death, loss to follow-up, or reaching the study time limit.

[0038] Training set and validation set: The data related to the subjects in Hospital A were used as the model training set, which included complete baseline characteristics and follow-up records. The data related to the subjects in Hospitals B and C were used as independent validation sets to test the generalization performance and stability of the model.

[0039] like Figure 2 As shown, the present invention provides a risk prediction system for type 2 diabetic nephropathy, which is applied to a method for predicting the risk of type 2 diabetic nephropathy. The system includes: The feature selection unit is used to analyze the influence degree of the factors affecting the onset of type 2 diabetic nephropathy using a multi-factor Logistic regression model, and to screen out the characteristic factors that have a significant impact on the onset of type 2 diabetic nephropathy; The feature preprocessing unit is used to perform principal component analysis on each feature factor to obtain the contribution of the feature factor; based on the contribution of the feature factor, determine the feature weight for centralizing the feature factor; based on the contribution of the feature factor, determine the feature weight for decentralizing the feature factor; and perform a game between the feature weight after decentralized processing and the feature weight after centralized processing to obtain the optimal feature weight; The deep learning unit is used to construct a risk prediction model for type 2 diabetic nephropathy based on feature factors weighted by optimal feature weights.

[0040] The present invention utilizes a neural network to construct a prediction model that can efficiently identify high-risk populations, thereby predicting the risk of type 2 diabetic nephropathy in patients by obtaining characteristic factors from clinical data, thereby achieving objective, accurate and standardized risk prediction.

[0041] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A method for predicting the risk of developing type 2 diabetic nephropathy, characterized in that: The following steps are involved: The multivariate Logistic regression model was used to analyze the influencing factors of type 2 diabetic nephropathy and screen out the characteristic factors that have a significant impact on the onset of type 2 diabetic nephropathy. Conduct principal component analysis on each characteristic factor to obtain the contribution of the characteristic factor; Based on the contribution of the characteristic factors, determining characteristic weights for centralizing the characteristic factors; Determining feature weights for decentralizing the feature factors based on the contribution of the feature factors; A game is played between the feature weights after decentralized processing and the feature weights after centralized processing to obtain the optimal feature weights; Based on the characteristic factors weighted by the optimal characteristic weights, a risk prediction model for type 2 diabetic nephropathy was constructed.

2. The method for predicting the risk of developing type 2 diabetic nephropathy according to claim 1, characterized in that: The influencing factors include: demographic characteristics, clinical information, laboratory tests, renal puncture pathology results, physical examination, lifestyle, medication use, past medical history and family history, socioeconomic status, and emotional state.

3. The method for predicting the risk of developing type 2 diabetic nephropathy according to claim 2, characterized in that: The screening method of the characteristic factors includes: SPSS26.0 and RStudio4.2.2 software were used to analyze the case data of type 2 diabetic nephropathy. The quantitative data that conformed to the normal distribution were expressed as mean ± standard deviation, and the qualitative data that did not conform to the normal distribution were expressed as frequency and percentage. The differences in influencing factors between patients without type 2 diabetic nephropathy and patients with type 2 diabetic nephropathy were compared. The t-test was used to compare the quantitative data with normal distribution between the groups, and the chi-square test and rank sum test were used to compare the qualitative data that did not conform to the normal distribution between the groups. The difference was statistically significant; Logistic regression analysis was used to screen the influencing factors, with the test level α=0.05, and the screened influencing factors were used as characteristic factors.

4. The method for predicting the risk of developing type 2 diabetic nephropathy according to claim 3, characterized in that: Methods for determining feature weights after centralization include: The contribution of the characteristic factor is normalized and used as the characteristic weight after the characteristic factor is centralized. The normalization process of the feature weights after the centralization process is: ; In the formula, After centralized processing i The feature weight of each feature factor, For the i The contribution of the characteristic factors, is the contribution set of each characteristic factor, for The minimum value of The maximum value in .

5. The method for predicting the risk of developing type 2 diabetic nephropathy according to claim 4, characterized in that: Methods for determining feature weights after decentralized processing include: The contribution of the characteristic factor is processed inversely, and the contribution after the inverse processing is normalized as the characteristic weight of the characteristic factor after decentralized processing; The inverse processing and normalization processing of the feature weight after the centralization processing are: ; In the formula, After decentralization i The feature weight of each feature factor, For the i The contribution of the characteristic factors, is the inverse set of contribution of each characteristic factor, The minimum value in The maximum value in , n is the total number of characteristic factors, min is the minimum operator, and max is the maximum operator.

6. The method for predicting the risk of developing type 2 diabetic nephropathy according to claim 5, characterized in that: The method for determining the optimal feature weight includes: The game theory centralized model is used to optimize the feature weights after decentralized processing and the feature weights after centralized processing. The decision model of the optimal feature weights is derived from the optimization goal of minimizing the weight deviation. The decision model of the optimal feature weights is: ; In the formula, F is the objective function value, min is the minimum operator, x is the identifier of w and v, for The linear coefficient of , T is the transposition operator; The decision model is transformed and solved using the optimal first-order derivative condition, and the solution equation for the optimal feature weight is obtained as follows: ; Solve the solution equation and normalize it to obtain the optimal feature weight, which is: ; ; In the formula, is the set of optimal feature weights of each feature factor, After normalization , is a set of linear coefficients, for The minimum value in The maximum value in .

7. The method for predicting the risk of developing type 2 diabetic nephropathy according to claim 6, characterized in that: The method for constructing the risk prediction model for type 2 diabetic nephropathy comprises: Construct training and test sets based on case data of type 2 diabetic nephropathy; The feature factors weighted by the optimal feature weights are used as the input of the convolutional neural network, and the risk score is used as the output of the convolutional neural network; The convolutional neural network was trained using the training set to obtain a risk prediction model for type 2 diabetic nephropathy, where the risk prediction model for type 2 diabetic nephropathy is: ; In the formula, Score is the risk score are the optimal feature weights of the 1st, 2nd, and nth feature factors respectively, They are the 1st, 2nd, and nth feature factors respectively, and CNN is a convolutional neural network; The test set was used to evaluate the performance of the risk prediction model for type 2 diabetic nephropathy.

8. The method for predicting the risk of developing type 2 diabetic nephropathy according to claim 7, characterized in that: In the process of training the convolutional neural network, a multi-dimensional cross-validation method is used to optimize the hyperparameters of the convolutional neural network; In the performance evaluation of the risk prediction model for type 2 diabetic nephropathy, the calibration curve and D'Agostino-Nam test were used to evaluate the calibration of the model, and the area under the ROC curve and Harrell's C index were used to evaluate the discrimination ability of the model.

9. The method for predicting the risk of developing type 2 diabetic nephropathy according to claim 7, characterized in that: The risk level determination method comprises: The risk score range of the top 25% of all cases in the case data of type 2 diabetic nephropathy was classified as the threshold range of high risk level, the risk score range of the middle 50% was classified as the threshold range of medium risk level, and the risk score range of the bottom 25% was classified as the threshold range of low risk level.

10. A system for predicting the risk of developing type 2 diabetic nephropathy, characterized in that: A method for predicting the risk of type 2 diabetic nephropathy according to any one of claims 1 to 9, the system comprising: The feature selection unit is used to analyze the influence degree of the factors affecting the onset of type 2 diabetic nephropathy using a multi-factor Logistic regression model, and to screen out the characteristic factors that have a significant impact on the onset of type 2 diabetic nephropathy; A feature preprocessing unit is used to perform principal component analysis on each feature factor to obtain the contribution of the feature factor; based on the contribution of the feature factor, determine the feature weight for centralizing the feature factor; based on the contribution of the feature factor, determine the feature weight for decentralizing the feature factor; and perform a game between the feature weight after decentralized processing and the feature weight after centralized processing to obtain the optimal feature weight; The deep learning unit is used to construct a risk prediction model for type 2 diabetic nephropathy based on feature factors weighted by optimal feature weights.

Images